Trust
Security at Semantiks
Effective date: August 5, 2026 · Last updated: November 18, 2026
This overview explains our security programme and how to obtain evidence for a review. Customer agreements, the applicable Data Processing Addendum (DPA), and agreed security exhibits govern contractual commitments. This overview does not reduce existing customer protections.
Certification status. Semantiks is pursuing ISO/IEC 27001:2022 certification. An active programme or planned audit is not a certificate. We do not represent that Semantiks holds ISO certification or a SOC 2 report. Security documents in our Trust Center are subject to access approval and, where applicable, confidentiality.
1. Compliance programme and independent assurance
Our ISO/IEC 27001:2022 programme has been active since July 20, 2026, supported by Vanta for control monitoring and evidence collection. Available audit and penetration-test evidence, its date, scope, and limitations can be requested under confidentiality. Completed assurance artifacts are shared when available. No certification date or outcome is guaranteed.
2. Hosting, provider processing, and personnel access
Primary application infrastructure is hosted on Google Cloud Platform, and the primary application database uses Supabase on Amazon Web Services, in United States regions. Hosting location does not by itself establish where every model, observability, or other provider processes data. Provider-specific locations and safeguards are described in the applicable DPA, subprocessor register, and agreed residency terms. Existing data-residency commitments continue to apply.
Personnel are individual contractors engaged directly by Semantiks Inc. Remote access from countries where they work is assessed separately from hosting location and is subject to confidentiality and access controls. The separate Mexican operations entity, Semtiks, is not currently appointed to process Customer Data for the Services.
3. Encryption
Our security programme requires encryption in transit and at rest appropriate to the system and data. Primary GCP and Supabase services provide platform encryption. System-specific configurations, including backup encryption, are described in the applicable security documentation. Binding security measures are described in the applicable DPA and security exhibit.
4. Artificial-intelligence model providers
Customer workflows may use OpenAI for inference and embeddings and Cohere for reranking. The applicable providers and processing activities are documented in the subprocessor register. Anthropic is used for internal engineering and is not a Customer-facing inference provider.
Semantiks does not use Customer Data to train generalized AI models and requires the corresponding contractual protection from model-provider subprocessors. OpenAI-specific modified-retention arrangements are conditional on verified activation and eligible accounts, projects, models, endpoints, and input types. They do not establish zero retention for other providers or for application logs and conversation traces. Provider-specific retention and any limitations must be reviewed in the DPA and subprocessor documentation.
5. Access to production systems
Access is granted according to role and business need, subject to confidentiality obligations, review, and an offboarding process. Our security policies require timely access removal and multi-factor authentication on applicable systems. Service-specific enforcement, review evidence, and exceptions can be requested during a security review. Customer-specific access obligations remain binding.
6. Secure development
Our development programme includes code review, production-branch protections, dependency scanning, and vulnerability remediation. Coverage and exceptions are assessed by repository and deployment path. Security reviewers may request the applicable control evidence.
7. Human oversight and sensitive data
Customer-configured agents support conversation supervision, escalation, and human takeover. Customers are responsible for selecting appropriate human review and testing for their use case, especially where outcomes may have legal or significant effects. AI output may be inaccurate and should be evaluated before use.
Customers should not assume automated detection, redaction, or masking of sensitive data. Conversations and execution traces can contain personal data and may be transmitted to the applicable model and observability providers. The DPA describes processing limitations and additional safeguards required for sensitive-data use cases.
8. Subprocessors
The subprocessor register identifies providers, activities, and processing locations and is available on request under confidentiality. The DPA governs authorization, advance change notice, objections, and Semantiks’ responsibility for appointed subprocessors. Selecting or enabling a provider does not by itself determine whether it is an independent Customer provider or a Semantiks subprocessor.
9. Retention and deletion
Customers may request export or deletion through Semantiks support, including during the service term. Unless earlier deletion is instructed or required, the standard post-termination retrieval period is 30 days, followed by deletion or de-identification of active copies without undue delay and no later than 60 days after termination. Earlier binding deadlines continue to apply. Residual backups remain protected and are subject to the applicable agreement, including the DPA’s seven-day database-backup commitment. These obligations do not imply a self-service deletion feature. See the Privacy Notice and DPA for details and legally required retention.
10. Security incidents
The DPA requires notification without undue delay after awareness of a Personal Data Breach, with preliminary notice within 24 hours and a fuller report within 72 hours of awareness, supplemented as the investigation progresses. Notification does not wait for the final forensic report. Applicable law and existing Customer-specific commitments continue to govern.
11. Availability
Service information is available on our status page. The status page provides operational updates; it does not create an uptime guarantee. Availability commitments, where offered, are set in the applicable Customer agreement.
12. Reporting a security issue
Report suspected vulnerabilities or security concerns to security@semantiks.ai. We acknowledge reports and work with you on validation and remediation. We do not pursue legal action against researchers who report in good faith, avoid privacy violations and service degradation, and give us reasonable time to remediate before disclosure.
13. Related documents
- Terms of Service
- Privacy Notice
- Trust Center — access approval may be required
- Data Processing Addendum — available on request, including before agreement acceptance, at hola@semantiks.ai
- Security measures and subprocessor register — available under confidentiality on request